Proceedings of the

Annual Stability Conference

Structural Stability Research Council
Atlanta, Georgia, April 21-24, 2020

Experimental investigation on cold-formed steel stiffened lipped channel

columns undergoing local-distortional interaction
Man-Tai Chen1, Ben Young2, André Dias Martins3, Dinar Camotim4, Pedro Borges Dinis5

Abstract
The objective of this work is to report a careful experimental investigation, planned at the University
of Lisbon and carried out at The University of Hong Kong, concerning the behaviour and ultimate
strength of cold-formed steel web-flange-stiffened (WFSLC) and web-stiffened (WSLC) lipped
channel columns undergoing local-distortional (L-D) interaction. It involves 31 specimens (16
WFSLC+ 15 WSLC), brake-pressed from high-strength zinc-coated grades G450, G500 and G550
structural steel sheets, exhibiting critical distortional-to-local buckling load ratios ranging between
0.75 and 1.88. The column geometries were carefully selected to enable testing fixed-ended columns
undergoing true L-D interaction and secondary distortional-bifurcation L-D interaction (ensuring
evidence of the latter required selecting rather slender columns) − all tested specimens exhibited the
expected L-D interactive failures. The specimen material properties were obtained from tensile
coupon tests and their initial geometrical imperfections were measured prior to testing. The
experimental results presented and discussed consist of column (i) load-displacement equilibrium
paths, (ii) photos evidencing the evolution of the column deformed configurations along those paths
(including the failure mode) and (iii) failure loads. Finally, the experimental failure load data
obtained are compared with their estimates provided by the currently codified DSM design
approaches for columns failing in L and D modes, showing their inadequacy to handle L-D
interactive failures − the fresh light shed by this comparison will contribute to the timely codification,
in the near future, of a DSM design procedure for columns affected by L-D interaction.

1. Introduction
It is well known that thin-walled open-section cold-formed steel (CFS) columns are highly susceptible
to several instability phenomena, involving either local (L), distortional (D) or global (G) buckling,
or any coupling between them, i.e., L-D, L-G, D-G or L-D-G interaction – see the recent state-of-the-
art report concerning CFS members affected by interactive buckling phenomena (Camotim et al.
2018). This investigation deals exclusively with columns affected by L-D interaction and aims at
providing additional experimental failure loads associated with collapses in L-D interactive modes.

1
Assistant Professor, Shanghai Jiao Tong University, <mantai.chen@sjtu.edu.cn> (formerly: Postdoctoral Researcher,
The University of Hong Kong)
2
Professor, The Hong Kong Polytechnic University, ben.young@polyu.edu.hk (formerly: Professor, The University
of Hong Kong)
3
Postdoctoral Researcher, Instituto Superior Técnico, University of Lisbon, <andrerdmartins@tecnico.ulisboa.pt>
4
Professor, Instituto Superior Técnico, University of Lisbon, <dcamotim@civil.ist.utl.pt>
5
Assistant Professor, Instituto Superior Técnico, University of Lisbon, <dinis@civil.ist.utl.pt>
A significant amount of work has been devoted to investigate the behavioural features inherent to this
relevant coupling phenomenon in short-to-intermediate columns. Indeed, such investigations deal with
(i) shell finite element analysis (e.g., Martins et al. 2015, Dinis & Camotim 2015), (ii) Generalised
Beam Theory (GBT)-based investigations used to assess the mechanics underlying this coupling
phenomenon (e.g., Martins et al. 2018), (iii) design proposals (e.g., Silvestre et al. 2012, Martins et al.
2017a, Matsubara et al. 2019), mostly based on the Direct Strength Method (DSM) concepts (Schafer
2008, Camotim et al. 2016, Schafer 2019), and (iv) experimental investigations (discussed next).
Although a significant number of experimental results concerning the behaviour and strength of CFS
columns affected by local-distortional (L-D) interaction is available in the literature, most of them dealing
with lipped channels, specimens providing clear experimental evidence of this coupling phenomenon and
ensuing failure load erosion are relatively scarce – certainly much less than those collected to propose,
calibrate and validate the existing local, distortional, global and local-global DSM design curves. To the
authors’ best knowledge, the available experimental results evidencing the occurrence of L-D interaction
in fixed-ended CFS columns are due to (i) Kwon & Hancock (1992), Young & Rasmussen (1998),
Kwon et al. (2009), Loughlan et al. (2012), Young et al. (2013), Chen et al. (2019) and Matsubara et al.
(2019), for lipped channel (LC) columns, (ii) Kwon et al. (2005), for LC and hat-section (H) columns,
(iii) Chen et al. (2019), for zed-section (Z) columns, (iv) Dinis et al. (2014) and Chen et al. (2019), for
rack-section (R) columns, (v) Chen et al. (2019), for channel columns with “return lips” (RLC), (vi)
Kwon & Hancock (1992), Kwon et al. (2009), Yap & Hancock (2011), He et al. (2014) and Chen et al.
(2019), for web-stiffened lipped channel (WSLC) columns, (vii) Yang & Hancock (2004), for
web/flange-stiffened lipped channel (WFSLC) columns, and (viii) Yap & Hancock (2008), for columns
with complex-stiffened cross-sections, totalling 134 test results. Naturally, reliable and illuminating
experimental data are essential to establish efficient DSM-based design rules to handle column L-D
interactive failures, which are still lacking in the current CFS specifications worldwide. Aiming to fill
this gap, Martins et al. (2017a) recently proposed a well-founded preliminary version of the sought
DSM-based design approach. It just needs to be “fine-tuned”, a task that (i) requires carefully obtained
fresh experimental failure loads, namely those reported in this paper, and (ii) should pave the way for
codification − the final goal of this research effort.
The main objective of this work is to present a careful experimental investigation, planned at the
University of Lisbon and carried out at The University of Hong Kong, concerning the behaviour and
ultimate strength of CFS WFSLC and WSLC columns undergoing L-D interaction. It comprises 31
specimens (16 WFSLC and 15 WSLC) (i) brake-pressed from high-strength zinc-coated structural steel
sheets of grades G450, G500 and G550, and (ii) exhibiting distortional-to-local critical buckling load
ratios ranging between 0.8 and 1.9. The column geometries selected, discussed in Section 2, were
carefully chosen to enable the testing of fixed-ended columns undergoing (i) “true L-D interaction”, i.e.,
with nearly coincident critical local and distortional buckling loads, and (ii) “secondary distortional-
bifurcation L-D interaction”, i.e., with the non-critical (distortional) buckling load visibly above the
critical (local) one. Acquiring clear evidence of the latter L-D interaction required selecting rather slender
columns – their measured dimensions are presented in Section 3.1. The specimen material properties
were obtained from tensile coupon tests (Section 3.2) and their initial geometrical imperfections were
measured prior to testing (Section 3.4). After addressing the set-up procedure, in Section 3.3, the
experimental results are presented in Section 3.5 and consist of column (i) load-displacement
equilibrium paths, (ii) deformed configuration evolutions along those paths, including the failure
mode, and (iii) failure loads. Finally, in Section 4, the experimental failure load data obtained are compared
with their estimates yielded by the currently codified DSM design curves for columns failing in pure L or D
modes, providing additional clear experimental evidence that these design curves are unable to handle

2
adequately L-D interactive failures. Moreover, the fresh light shed by this comparison will contribute to the
timely codification of an efficient (safe, accurate and reliable) DSM design procedure in the near future.

2. Column geometry selection

The success of any experimental investigation is directly linked to its planning. In the present case,
the selection of the specimen geometry plays a pivotal role in achieving the objectives proposed. The
aim of this test campaign is to obtain experimental failure loads of fixed-ended columns exhibiting
WSLC and WFSLC cross-sections, and having distortional-to-local critical buckling load ratios
(RDL=PcrD/PcrL) varying between 0.8 and 2.0 (approximately), i.e., prone to either (i) “true L-D
interaction” (0.8<RDL<1.3) 6 or (ii) “secondary distortional-bifurcation L-D interaction” (RDL>1.3)
(Martins et al. 2015). Note that, columns having RDL<1.3 are prone to distortional or “secondary
local-bifurcation L-D interaction” − since the latter are characterised by a negligible failure load
erosion (i.e., are adequately handled by the current DSM distortional design curve), they fall outside
the scope of this investigation. In order to reach the sought column specimen geometries, their
selection procedure should:
(i) provide specimens characterised by 0.8<RDL<2.0 (as previously discussed);
(ii) be associated with critical global buckling loads (PcrG) much higher than their PcrD and PcrL
counterparts (to avoid the interaction with the global buckling mode, i.e., L-D-G interaction), and
(iii) identify columns prone to “secondary distortional-bifurcation L-D interaction” that are (locally)
slender, since stocky columns with RDL>1.3 are prone to local failure and exhibit no interaction.
On the other hand, the column selection procedure is “constrained” by the following limitations:
(i) specimens with lengths not exceeding L=3000mm (testing machine limitation);
(ii) specimens without small lip and stiffeners widths (manufacturing limitation);
(iii) thickness values and steel grades of t=1.0mm (G550), t=1.2mm (G500) or t=1.5mm (G450), and
(iv) failure loads (PExp) not exceeding 220kN (load actuator capacity) – to ensure this condition,
materially and geometrically non-linear shell finite element analyses (not addressed here) were
conducted and showed that all the selected specimen have expected failure loads well below the
above limit value (as presented in Section 3.5.2).
The column selection procedure was conducted through and “trial-and-error” approach by means of
the GBT-based code GBTUL (Bebiano et al. 2018), mainly due to its computational efficiency and
modal nature – note that, for the sake of simplicity, the round corners were neglected since they are
typically lower than the wall thickness. In addition, the Young’s modulus and the Poisson’s ratio
were considered equal to 210GPa and 0.30, respectively. Table 1 shows the output of this selection
procedure and displays the obtained cross-section dimensions (D, B, bl, t – see Fig. 1, and length L),
critical (global, local and distortional) buckling loads, critical distortional half-wave number
(indicated between brackets in the column PcrD), RDL value and local (λL) and distortional slenderness
(λD) – the v-shape (45º) stiffener widths bws and bfs (see Fig. 1) were considered equal to 152mm in

6
At this stage, it should be mentioned that these RDL values, providing the “boundaries” between “true L-D interaction” and
“secondary local or distortional-bifurcation L-D interaction”, are only approximate and somewhat arbitrary − they are just
intended to identify, with reasonable accuracy, the columns undergoing strong L-D interaction. These values were first
proposed by the authors (Martins et al. 2015), in the context of lipped channel, hat-section, zed-section and rack-section
columns. Subsequently, it was found that they apply also to web-stiffened (Martins et al. 2016) and web-flange-stiffened
(Martins et al. 2017b) lipped channel columns, even if this information is not provided explicitly in the above references.

3
Table 1: Selected column specimen (i) geometries, (ii) critical local, distortional and global buckling loads, and (iii) local
and distortional slenderness
Specimen D B bl t L Steel fy PcrG PcrL PcrD RDL λL λD
(mm) (mm) (mm) (mm) (mm) Grade (MPa) (kN) (kN) (kN) (-) (-) (-)
WS-1 280 150 17 1.5 2500 G450 530 1776 86.0 71.2(2) 0.83 2.41 2.64
WS-2 280 150 17 1.5 2200 G450 530 2293 86.2 77.7(1) 0.90 2.40 2.53
WS-3 220 130 17 1.5 1700 G450 530 2181 98.4 95.5(1) 0.97 2.06 2.09
WS-4 220 130 17 1.5 1500 G450 530 2801 98.3 100.0(1) 1.02 2.06 2.05
WS-5 260 150 17 1.5 1800 G450 530 3005 84.3 81.8(1) 0.97 2.39 2.43
WS-6 260 150 17 1.5 1550 G450 530 4052 84.2 88.6(1) 1.05 2.39 2.33
WS-7 270 160 17 1.5 1500 G450 530 4928 77.8 86.9(1) 1.12 2.55 2.41
WS-8 270 160 17 1.5 1650 G450 530 4073 77.7 81.3(1) 1.05 2.55 2.49
WS-9 270 170 17 1.2 1650 G500 560 3447 36.7 53.3(1) 1.45 3.47 2.88
WS-10 270 170 17 1.2 1950 G500 560 2468 36.7 47.0(1) 1.28 3.47 3.06
WS-11 250 150 17 1.0 1750 G550 600 1973 24.8 40.7(1) 1.64 3.47 2.96
WS-12 250 150 17 1.0 1650 G550 600 2219 24.8 42.7(1) 1.72 3.80 2.89
WS-13 250 150 17 1.0 1350 G550 600 3315 24.8 50.7(1) 2.04 3.80 2.66
WFS-1 250 200 17 1.0 1750 G550 600 2431 63.2 51.9(1) 0.82 2.62 2.89
WFS-2 250 200 17 1.0 1650 G550 600 2735 63.2 56.8(1) 0.90 2.62 2.76
WFS-3 250 200 17 1.0 1500 G550 600 3309 63.7 66.1(1) 1.04 2.61 2.56
WFS-4 250 220 17 1.2 1250 G500 560 6159 113.2 108.7(1) 0.96 2.20 2.25
WFS-5 270 240 17 1.2 1300 G500 560 7162 102.6 101.3(1) 0.99 2.40 2.42
WFS-6 270 180 17 1.0 1700 G550 600 2749 53.2 56.4(1) 1.06 2.81 2.73
WFS-7 280 190 20 1.0 2000 G550 600 2298 51.5 54.9(1) 1.07 2.93 2.84
WFS-8 250 220 17 1.0 1350 G550 600 4400 65.6 76.8(1) 1.17 2.64 2.44
WFS-9 270 180 17 1.0 1400 G550 600 4054 53.4 76.0(1) 1.42 2.81 2.35
WFS-10 270 180 17 1.0 1500 G550 600 3531 53.4 68.2(1) 1.28 2.81 2.48
WFS-11 280 190 20 1.0 1500 G550 600 4085 51.7 82.4(1) 1.59 2.92 2.32
WFS-12 300 200 20 1.0 1650 G550 600 4042 47.0 71.3(1) 1.52 3.15 2.56
WFS-13 300 200 20 1.0 1500 G550 600 4891 46.9 80.9(1) 1.72 3.15 2.40

all specimens. The close observation of this table shows that (i) all the aforementioned requirements
are satisfied and (ii) a total of 26 (13 WFSLC+ 13 WSLC) specimens were selected. Indeed, there are
(i) 14 specimens with t=1.0mm (grade G550), labelled WS-11, 12, 13 and WFS-1, 2, 3, 6, 7, 8, 9, 10,
11, 12, 13, (ii) 4 specimens with t=1.2mm (grade G500), labelled WS-9, 10, and WFS-4, 5, and (iii) 8
specimens with t=1.5mm (grade G450), labelled WS-1, 2, 3, 4, 5, 6, 7, 8 and (iv) only one specimen
with a critical distortional buckling mode exhibiting two half-waves (WS-1). For illustrative
purposes, Fig. 2(a)-(c) shows the critical local (1) and distortional (2) buckling mode shapes of
columns WS-1, WS-13 and WFS-5, respectively – note that local buckling is triggered by the web
and flanges, respectively in all WSLC and WFSLC columns.

(a) (b)
Figure 1: Cross-section shapes and dimensions of (a) web-stiffened, and (b) web-flange-stiffened lipped channels.

4
 (a1) (a2) (b1) (b2) (c1) (c2)
Figure 2: Critical (1) local and (2) distortional buckling mode shapes of columns (a) WS-1, (b) WS-13 and (c) WFS-5.

3. Experimental investigation
3.1 Test specimens
The CFS test specimens were brake-pressed from high strength zinc-coated grades G450, G500 and
G550 structural steel sheets having nominal yield stresses of 450, 500 and 550MPa, respectively. The ends
of the column specimens were welded to 16mm thick steel end plates, to ensure full contact between the
specimen end cross-sections and bearing plates. The base metal thickness (t*) was measured by removing
the zinc coating by acid-etching. The thickness of the zinc coating, measured on one side of the cross-
section, was (i) 23.5/21.5/26.0m for wall thickness 1.0/1.2/1.5mm (WSLC specimens) and (ii)
30.5/22.5m for wall thickness 1.0/1.2mm (WFSLC specimens). The measured cross-section dimensions,
cross-section area (A) and length (L) for each specimen are shown in Table 2 (definitions/nomenclature
given in Fig. 1). Note that, in order to assess the reliability of the test procedure, five pairs of specimens
with nearly coincident dimensions were considered, namely WS-5+5R, WS-6+6R, WFS-5+5R,
WFS-6+6R and WFS-9+9R (see Table 2).

3.2 Material properties

Tensile coupon tests were conducted to determine the material properties of the CFS WSLC and WFSLC
column specimens. The coupons were extracted, in the longitudinal direction, from the central part of the
unstiffened portions of the webs of specimens WFS-4, WFS-8, WS-4, WS-9 and WS-13, covering all the
distinct wall thickness and cross-section shape considered in this investigation. The column specimens
with the same nominal wall thickness and steel grade were fabricated from the same batch of steel. For
instance, the material properties obtained from specimen WFS-4 (t1.20mm) were deemed similar to
those of the other specimens sharing the same wall thickness and steel grade, i.e., specimens WFS-5 and
WFS-5R (see Table 1) – the same “extrapolation” was made for all the other specimens.
The coupon dimensions were determined according to the American standard ASTM-E8M (2016)
(tensile testing of metals) using 12.5mm wide coupons of gauge length 50mm. An MTS displacement-
controlled testing machine using friction grips was used to conduct the coupon tests. In addition, the
coupons were tested according to the recommendations of Huang & Young (2014) for cold-formed
carbon steel. A calibrated extensometer of 50 mm gauge length was used to measure the coupon
specimen longitudinal strains. A data acquisition system was used to record the load and the strain
readings at regular intervals during the tests. The tests were paused for 100 seconds near the 0.2% proof
stress and the ultimate tensile strength, to allow for the stress relaxation associated with plastic straining

5
Table 2: Measured cross-section dimensions, lengths and cross-section area of the tested specimens (WSLC and WFSLC).
Specimen D B bl bf bw bfs bws t t* ri L A
(mm) (mm) (mm) (mm) (mm) (mm) (mm) (mm) (mm) (mm) (mm) (mm2)
WS-1 279.6 150.5 15.7 - 122.0 - 18.6 1.565 1.513 0.97 2501 969.9
WS-2 278.7 151.1 16.1 - 121.7 - 19.2 1.567 1.515 0.72 2201 973.2
WS-3 220.6 129.8 15.5 - 92.2 - 18.2 1.565 1.513 0.59 1701 807.1
WS-4 222.3 129.6 16.0 - 93.8 - 18.4 1.574 1.522 0.96 1502 821.3
WS-5 261.9 150.3 16.4 - 113.3 - 18.5 1.564 1.512 0.72 1803 941.5
WS-5R 262.7 149.9 16.3 - 113.2 - 18.7 1.572 1.520 0.84 1801 946.1
WS-6 258.8 150.6 16.3 - 111.4 - 18.8 1.565 1.513 0.72 1552 937.5
WS-6R 259.5 149.8 15.6 - 112.3 - 17.6 1.560 1.508 0.60 1554 927.4
WS-7 269.1 160.3 16.0 - 116.8 - 18.3 1.553 1.501 0.72 1500 974.8
WS-8 269.0 160.4 15.7 - 117.3 - 18.2 1.563 1.511 0.72 1651 981.8
WS-9 274.5 170.4 16.7 - 120.4 - 18.5 1.258 1.215 0.87 1653 826.9
WS-10 272.7 171.6 16.4 - 120.2 - 18.4 1.262 1.219 0.74 1955 830.4
WS-11 252.1 151.2 16.5 - 109.3 - 19.2 1.060 1.013 0.97 1753 634.0
WS-12 253.3 151.6 16.3 - 109.1 - 19.3 1.058 1.011 0.85 1652 632.5
WS-13 255.5 150.9 15.8 - 109.8 - 18.7 1.059 1.012 0.85 1350 630.5
WFS-1 252.4 202.2 15.9 86.7 108.5 19.4 19.0 1.052 0.991 0.72 1752 755.9
WFS-2 252.5 201.5 15.8 83.8 108.6 19.3 19.8 1.064 1.003 0.97 1652 770.5
WFS-3 252.2 201.6 15.9 83.5 108.2 19.3 18.9 1.063 1.002 0.72 1503 762.3
WFS-4 250.1 220.5 15.8 93.7 107.7 18.8 18.4 1.254 1.209 1.00 1254 951.9
WFS-5 271.4 241.2 15.0 103.2 117.2 18.6 18.6 1.263 1.218 0.62 1299 1020.8
WFS-5R 273.8 240.2 15.0 103.3 119.0 18.6 18.1 1.265 1.220 0.62 1299 1026.6
WFS-6 270.6 182.6 16.2 73.4 117.3 18.6 18.8 1.067 1.006 0.72 1702 738.5
WFS-6R 273.3 181.4 15.8 73.7 119.2 19.6 19.4 1.070 1.009 0.72 1703 750.8
WFS-7 283.8 193.2 17.2 78.0 123.7 18.6 18.5 1.055 0.994 0.72 2005 764.2
WFS-8 252.5 222.0 15.5 94.2 109.4 19.1 16.3 1.059 0.998 0.72 1355 799.6
WFS-9 273.0 182.7 17.7 73.4 119.1 19.2 18.4 1.055 0.994 0.72 1398 738.9
WFS-9R 271.5 182.1 16.4 73.6 118.8 19.5 19.8 1.071 1.010 0.84 1401 753.6
WFS-10 270.8 182.6 15.6 74.8 117.2 19.5 19.9 1.059 0.998 0.72 1504 743.5
WFS-11 282.6 191.6 17.5 78.4 122.8 19.0 20.4 1.061 1.000 0.72 1501 774.9
WFS-12 300.5 202.0 16.9 82.9 128.4 19.9 20.6 1.060 0.999 0.72 1654 818.2
WFS-13 302.6 201.6 16.7 83.0 134.0 18.1 19.2 1.069 1.008 0.72 1502 816.0

and to obtain the static stress-strain relationship. The nominal and measured material properties
obtained from the coupon tests are summarised in Table 3 and include the measured values of (i)
Young’s modulus (E), (ii) static 0.2% proof stress (0.2), (iii) static tensile strength (u) and
corresponding strain (u), and (iv) elongation after rupture (f) – note that, the various values obtained
for the Young’s modulus are very similar. The constitutive laws obtained for all the coupons analysed
are very similar − for illustrative purposes, Fig. 3 shows the stress-strain (test and static) curves
obtained from one coupon test analysed (specimens WFS-4 and WS-13). Naturally, the numerical
analyses intended to simulate these tests must employ the appropriate true stress-strain curves, obtained
from the curves displayed in Figs. 3(a)-(b).
Table 3: Nominal and measured material properties of CFS WSLC and WFSLC specimens.
Test Series Nominal Measured
t 0.2 E 0.2 u u f
(mm) (MPa) (GPa) (MPa) (MPa) (%) (%)
WFS-4 1.2 500 215.2 617.2 631.4 5.1 9.1
WFS-8 1.0 550 215.6 601.7 615.4 5.0 8.3
WS-4 1.5 450 212.6 488.2 507.6 6.0 11.9
WS-9 1.2 500 214.1 612.2 627.6 5.0 8.1
WS-13 1.0 550 216.7 602.7 613.8 4.9 8.0

6
 (a) (b)
Figure 3: Tensile coupon tests: illustrative stress-strain curves concerning specimens (a) WFSLC-4 and (b) WSLC-13.

3.3 Test set-up and procedure

A schematic representation of the test rig and set-up employed in this investigation is shown in Fig.
4. It consists of a vertical testing frame conceived to conduct displacement-controlled tests, an
essential feature in assessing geometrically non-linear behaviours involving the need to obtain a
structural response after the peak load has been reached. A servo-controlled hydraulic testing
machine was employed to apply a compressive axial force to the column specimens (see the top of
Fig. 4), which can go up to 220kN. In addition, two thick steel plates were welded to the specimen
end cross-sections (see the top and bottom of Fig. 4). The specimen top end plate was bolted to the
rigid flat bearing plate, which was restrained against bending rotations about the minor and major
axes, twist rotations and warping. On the other hand, a special bearing device was used at the bottom
end support. Initially, this special bearing was free to rotate in any direction, since it contains a
spherical hinge. Then, the actuator ram was moved slowly until the specimen bottom end plate was
in full contact with the special bearing, following the imposition of a small initial load
(approximately 2kN). This procedure eliminated any possible gaps between the special bearing and
the specimen bottom end plate. Then, the bottom end plate was bolted to the special bearing, which
was subsequently restrained against bending rotations (minor and major-axis) and twisting, by means
of vertical and horizontal bolts, respectively – these bolts locked the bearing in position after full
contact was achieved. Thus, the use of this special bearing makes it possible to materialise a fixed-
ended support, as it restrains the end-section minor and major-axis bending rotations, twist rotations
and warping.

Displacement control was used to drive the hydraulic actuator at a constant speed of 0.2mm/min for
all column specimens. Moreover, in order to measure relevant displacements along the procedure,
several transducers (Linear Variable Displacement Transducers – LVDTs) were installed along the
specimen length – Figs. 5 and 6 show the transducer arrangement. Regardless of the cross-section
shape, three longitudinal displacement transducers (A1 to A3 in Fig. 6) were installed to measure the
column axial shortening. In addition, eleven (WSLC) or twelve (WFSLC) horizontal transducers
were installed to measure the deformations of the mid-height cross-section – note that the horizontal
transducers near the corners were located 15mm away from those same corners (e.g., web-flange
corners). In addition, in both specimens (see Figs. 6(a)-(b)), (i) three transducers were placed in each
stiffened or flat flange (T1 to T3 and T9 to T11), to capture the corresponding in-plane distortional
displacements, (ii) four transducers were placed in the stiffened web (T4, T5, T7 and T8), to capture
its in-plane distortional (T4, T7 and T8) and local (T4, T5 and T7) displacements. Note also that

7
either one (T6 − WFSLC specimens) or two (T6 and T12 − WSLC specimens) additional transducers
were placed 130mm away from the specimens mid-height (see Figs. 6(a)-(b)), in order to detect the
cases in which the maximum displacements did not occur at the specimen mid-height − 130mm is
close to the width of the flat web portions (bw). It should be mentioned that, rigorously speaking, the
horizontal transducers do not measure exclusively local and/or distortional displacements, since
significant minor-axis flexure emerges as loading progresses. It is due to the effective centroid shift
effects caused by the substantial cross-section stress redistribution occurring at advanced loading
stages (e.g., Young & Rasmussen 1999), a phenomenon that was recently quantified by means of
GBT-based post-buckling analysis in (Martins et al. 2018) − in this case, the stress redistribution
stems mainly from the distortional deformations. Lastly, it is worth noting that the characterisation of
the load-displacement equilibrium paths was achieved through a data acquisition system that
recorded the applied load and the readings of the displacement transducers at regular intervals during
the tests.

Actuator

End plate

Specimen

Transducer

End plate

Special bearing

Figure 4: Fixed-ended column test rig and set-up.

8
 (a) (b)
Figure 5: Transducer arrangement at the mid-height cross-section of the (a) WFSLC and (b) WSLC specimens.

(a) (b)
Figure 6: Location of the transducers at top and mid-height cross-sections (a) WFSLC and (b) WSLC specimens.

9
3.4 Initial geometrical imperfections
In order conduct future numerical simulations, initial geometrical imperfections were measured prior to
testing, namely distortional and global (minor-axis flexure displacements and torsional rotations). The
initial imperfection distortional component (D – see Fig. 7(a)) was obtained as half the difference
between the readings concerning (i) the web width D and (ii) the distance between the two flange-lip
corners D2 (D=(D2-D)/2) – this imperfection component was obtained on the basis of readings provided
by a theodolite at three locations: specimen top, bottom and mid-height cross-sections. The D values
obtained are given in Table 4, where the positive sign indicates “outward” (distortional) flange-lip
motions. The minor-axis flexural (F) and torsional (T) initial geometrical imperfections were based on
the readings indicated in Figs. 7(b)+(c), always measured at the specimen mid-height cross-section web-
flange corners. These initial imperfection components are also given in Table 4, “adequately” normalised,
where positive F/L and 2T/D values indicate, respectively, (i) minor-axis bending curvatures towards the
web, and (ii) clockwise torsional rotations. It is also worth mentioning that (i) only discrete displacements
were measured (mainly at mid-height), which means that no initial geometrical longitudinal profiles are
available, and that (ii) these specimens are rather slender (see/recall Section 2), which means that their failure
loads are not significantly affected by the initial geometrical amplitude and/or shape – see Martins et al. (2017a).

Table 4: Measured initial geometrical imperfections for all the specimens tested: distortional and minor-axis flexure
displacements, and torsional rotations (see nomenclature in Fig. 7)
Flexural Torsional
Specimen Distortional imperfection (mm), D = (D2 – D)/2
imperfection imperfection
Top end Mid-height Bottom end F /L 2T /D
WFS-1 7.00 12.50 2.25 -1/27375 1/181
WFS-2 6.50 12.25 7.50 1/51625 1/209
WFS-3 4.25 11.50 6.75 1/5917 1/496
WFS-4 -1.00 5.50 0.00 -1/3956 1/658
WFS-5 3.50 14.75 5.00 1/10228 1/1069
WFS-5R 0.00 8.50 1.50 0 1/269
WFS-6 7.75 11.50 4.50 1/13402 1/355
WFS-6R 8.00 13.50 6.00 0 1/359
WFS-7 0.00 2.00 4.75 -1/31328 1/172
WFS-8 2.50 10.50 3.25 1/105 1/10
WFS-9 5.66 12.30 7.44 0 1/269
WFS-9R -0.21 3.93 -0.48 1/4420 1/428
WFS-10 7.50 10.25 5.00 1/7874 1/305
WFS-11 6.75 10.50 4.25 1/5909 1/556
WFS-12 0.25 2.75 0.75 1/6512 0
WFS-13 -2.00 4.75 1.00 -1/23841 1/217
WS-1 10.00 7.50 13.25 1/2624 1/732
WS-2 2.75 -4.50 -2.50 1/2476 1/1097
WS-3 3.25 1.75 1.25 1/2232 0
WS-4 2.00 2.00 4.00 1/2626 1/351
WS-5 4.50 2.50 5.75 1/5166 1/317
WS-5R 4.50 4.00 8.25 1/2026 1/517
WS-6 3.75 1.25 2.50 1/2037 1/509
WS-6R -1.75 -4.50 -4.00 1/3059 0
WS-7 -0.25 -2.50 -0.50 1/2953 1/530
WS-8 -0.75 -1.25 1.75 1/8644 1/192
WS-9 -2.25 -4.50 -2.75 1/8654 1/722
WS-10 5.50 -1.25 -1.50 1/2566 1/268
WS-11 1.75 -0.75 4.00 1/1454 1/663
WS-12 4.50 0.25 0.75 1/2888 1/285
WS-13 -0.50 -1.25 4.00 1/2126 1/251

10
 (a) (b) (c)
Figure 7: Initial geometrical imperfection measurements at specimen mid-height cross-section: (a) distortional
displacements, (b) minor-axis flexure displacements, and (c) torsional rotation.

3.5 Test results

The experimental results obtained from this test campaign, presented and discussed in the two next sub-
sections, consist of (i) load-displacement equilibrium paths and photos showing deformed configurations
along those paths, including the failure mode (Section 3.5.1), and (ii) failure loads (Section 3.5.2). Before
these results are addressed, Figs. 8(a)-(o) (WSLC) and Figs. 9(a)-(p) (WFSLC) display the failure modes,
i.e., the deformed configurations at the onset of collapse, of all the 31 specimens tested. These photos
provide clear evidence of local and (mostly) distortional deformations taking place along the column,
which confirms that all these specimens are affected by L-D interaction (as intended). Moreover, since
most of the specimens tested exhibited similar structural responses, only a representative fraction of the
results obtained for the 31 columns are presented and discussed in this manuscript.

3.5.1 Deformed configurations and equilibrium paths

3.5.1.1 True L-D interaction
This sub-section presents and discusses the structural response of four column specimens with close
critical local and distortional buckling loads, namely specimens WS-1, WS-5, WFS-6 and WFS-9.
Figs. 10 (WS-5), 11 (WFS-6) and 12 (WFS-9) show several views of column deformed configurations (i)
along the equilibrium path ascending branch (before collapse – Figs. 10(a)-12(a)), (ii) at the onset of
collapse (Figs. 10(b)-12(b)) and (iii) along the equilibrium path descending branch (after collapse – Figs.
10(c)-11(c)) or after removing the applied load (Fig. 12(c)). Fig. 13, concerning column WS-1, shows the
deformed configurations prior to collapse (Fig. 13(a)) and at the onset of collapse (Fig. 13(b)). Moreover,
Fig. 14 shows recorded equilibrium paths, concerning the aforementioned columns, displaying the applied
load P, measured by the hydraulic actuator load cell, versus (i) the column axial shortening (average of the
measurements by transducers A1, A2 and A3 – Figs. 14(a1)-(a4)), and the mid-height (ii) flange-lip
vertical displacements (measured by transducer T1), (iii) web-top flange horizontal displacements
(measured by transducer T4) and (iv) mid-top web horizontal displacements (measured by transducer
T5). The observation of these results leads to the following comments:
(i) First of all, pronounced local and distortional deformations occur in the columns prior to collapse, as
shown in Figs. 10-13(a1)-(a4), thus attesting that these columns experience “true L-D interaction”.
Moreover, the columns selected (recall Section 2) were designed to exhibit high local and distortional
slenderness values, which allows for the development of such pronounced deformations, which
become even more visible at the onset of collapse (see Figs. 10-13(b1)-(b4)).
(ii) All column failure modes exhibit one or two distortional half-waves − same as the critical distortional
half-wave number (recall Section 2). Except column WS-1, which exhibited two distortional half-
waves, all the columns exhibit one distortional half-wave involving either “inward” or “outward”

11
 (a) (b) (c) (d)

(e) (f) (g) (h)

(i) (j) (k) (l)

(m) (n) (o)

Figure 8: L-D interactive failure modes of WSLC specimens (a)-(o) 1, 2, 3, 4, 5, 5R, 6, 6R, 7, 8, 9, 10, 11, 12 and 13.

12
 (a) (b) (c) (d)

(e) (f) (g) (h)

(i) (j) (k) (l)

(m) (n) (o) (p)

Figure 9: L-D interactive failure modes of WFSLC specimens (a)-(p) 1, 2, 3, 4, 5, 5R, 6, 6R, 7, 8, 9, 9R, 10, 11, 12 and 13.

13
 (a1) (a2) (a3) (a4)

(b1) (b2) (b3) (b4)

(c1) (c2) (c3) (c4)

Figure 10: WSLC-5 column deformed configurations (a) prior to collapse (ascending equilibrium path), (b) at the onset of collapse
(failure mode) and (c) after collapse (descending equilibrium path).

flange-lip motions. In fact, all WSLC columns (except the WS-1 one) exhibited “inward” flange-lip
motions (see Fig. 8), which is in agreement with the findings reported in (Martins et al. 2016),
where most initially imperfect columns tend to exhibit “inward” motions at fairly advanced post-
buckling stages – note that, depending on the initial geometrical imperfection pattern, these
columns may also exhibit “outward” flange-lip motions. On the other hand, all WFSLC
columns exhibited one distortional half-wave, in most cases involving “outward” flange-lip
motions, unlike their WSLC counterparts and now in close agreement with the findings reported in
(Martins et al. 2017b). Indeed, Fig. 9 shows that 14 (out of 16) columns exhibited “outward”
(distortional) flange-lip motions. As far as the local deformations are concerned, no relation can be

14
 (a1) (a2) (a3) (a4)

(b1) (b2) (b3) (b4)

(c1) (c2) (c3) (c4)

Figure 11: WFSLC-6 column deformed configurations (a) prior to collapse (equilibrium path ascending branch), (b) at the onset of
collapse (failure mode) and (c) after the collapse (equilibrium path descending branch).

established with the critical local buckling modes as was reported in (Martins et al. 2018) – since the
distortional deformations are always highly dominant, no information about the number of local
half-waves can be obtained from the measurement of (any) displacement longitudinal profile.
(iii) The plastic hinges appear either near the end regions or at mid-span, and are also in close agreement
with the numerical results reported in (Martins et al. 2016, 2017). The main difference between the
experimental and numerical features occurs in the columns with plastic hinges near the end regions:
two plastic hinges, equally distant from each end, are obtained numerically, while only one is
observed experimentally − this is most likely due to the non-symmetry exhibited by the initial
geometrical imperfections, unlike the critical-mode imperfections involved in the numerical simulations.

15
 (a1) (a2) (a3) (a4)

(b1) (b2) (b3) (b4)

(c1) (c2) (c3) (c4)

Figure 12: WFSLC-9 column deformed configurations (a) prior to collapse (equilibrium path ascending branch), (b) at the onset of
collapse (failure mode) and (c) after the collapse (equilibrium path descending branch) or after the applied load has been removed.

The location of each plastic hinge is clearly illustrated further ahead in the paper − see Figs. 18+19.
(iv) The column axial shortening grows almost linearly with the applied load in the early loading stages
only − as loading progresses, this growth becomes gradually more non-linear (see Figs. 14(a1)-(a4)).
(v) Confirming recent numerical findings (Martins et al. 2018), the distortional deformations play a
dominant role in the behaviour of all the columns tested, when compared with their local and global
counterparts. This fact is clearly illustrated and quantified in Figs. 14(b1)-(b4), providing the
displacements measured by transducer T1 (those measured by transducer T11 are very similar and,
thus, not shown here) – note that small portions of these displacements may be due to torsion present
in the initial imperfections.

16
 (a1) (a2) (a3) (a4)

(b1) (b2) (b3) (b4)

Figure 13: WSLC-1 column deformed configurations (a) prior to collapse (equilibrium path ascending branch) and (b) at
the onset of collapse (failure mode).

(vi) The displacements measured by transducer T4 consist overwhelmingly of contributions from minor-
axis flexure arising from the stress redistribution caused mostly by the distortional deformations
but also, to a lesser extent, by the local ones (Martins et al. 2018). Rigorously speaking, these
displacements may also contain tiny contributions from local and distortional deformations −
recall that T4 is not placed exactly at the web-flange corner. It is interesting to note that this
displacement points towards either the web (WSLC columns), as in plain lipped channels (Martins
et al. 2018), or the lips (WFSLC columns). Moreover, a substantial increase is observed when the
applied load reaches the critical buckling load level (see Fig. 14(c)). For applied loads below this
level, the minor-axis flexural displacements are very small and stem exclusively from the initial
geometrical imperfections.
(vii) As already mentioned in item (ii), the contribution from local deformations cannot generally be
quantified in members affected by L-D interaction, regardless of the measurement location, because
they always appear jointly with highly dominant distortional deformations. Indeed, the displacement
measured by transducer T5 combines contributions from distortional, local and minor-axis flexural
deformations (and possibly also a tiny contribution from the torsion present in the initially imperfect
geometry). Only a rigorous modal analysis is able to quantify the above contributions (e.g.,
(Martins et al. 2018)). Nonetheless, Figs. 10-13 provide fairly clear evidence of the presence of
local deformations, particularly in the flanges of WSLC columns or in the web of WFSLC columns.
(viii)Since the WS-1 column exhibits two distortional half-waves, unlike the remaining specimens, its
equilibrium paths (Figs. 14(a4)-(d4)) and deformed configurations (Figs. 13(a1)-(b4)) are quite distinct.

17
 WS-5 P (kN) WS-5 P (kN) WS-5 P (kN) WS-5
140 140 140

120 120 120

100 100 100

80 80 80

60 60 60

40 40 40

20 20 20

0 0 0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 0 10 20 30 40 50 0 5 10 15 20 0 5 10 15 20 25 30
Axial shortening (mm) Displacement T1 (mm) Displacement T4 (mm) Displacement T5 (mm)

(a1) (b1) (c1) (d1)

WFS-6 P (kN) WFS-6 P (kN) WFS-6 P (kN) WFS-6
100 100 100

80 80 80

60 60 60

40 40 40

20 20 20

0 0 0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 0 10 20 30 40 50 -0.5 0.0 0.5 1.0 1.5 2.0 2.5 0 5 10 15 20
Axial shortening (mm) Displacement T1 (mm) Displacement T4 (mm) Displacement T5 (mm)

(a2) (b2) (c2) (d2)

WFS-9 P (kN) WFS-9 P (kN) WFS-9 P (kN) WFS-9
100 100 100

80 80 80

60 60 60

40 40 40

20 20 20

0 0 0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 0 10 20 30 40 -1 0 1 2 3 4 5 6 7 0 5 10 15 20 25
Axial shortening (mm) Displacement T1 (mm) Displacement T4 (mm) Displacement T5 (mm)

(a3) (b3) (c3) (d3)

WS-1 P (kN) WS-1 P (kN) WS-1 P (kN) WS-1
120 120 120

100 100 100

80 80 80

60 60 60

40 40 40

20 20 20

0 0 0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 0 5 10 15 20 0 3 6 9 12 15 0 5 10 15 20
Axial shortening (mm) Displacement T1 (mm) Displacement T4 (mm) Displacement T5 (mm)
(a4) (b4) (c4) (d4)
Figure 14: Experimental equilibrium paths: applied load P vs. (a) axial shortening, and displacements (b) T1, (c) T4 and
(d) T5, for specimens (1) WS-5, (2) WFS-6, (3) WFS-9 and (4) WS-1.

18
 First of all, the presence of initial geometrical imperfections with a large amplitude (see Table 4)
leads to a highly non-linear behaviour since the early loading stages and until the applied load reaches
approximately 80kN (close to the critical buckling load level). Between 80kN and 100kN, the
displacement T1 remains practically constant, which is due to a fairly abrupt “configuration switch”
from one to two distortional half-waves: the bottom and top “half columns” exhibit “inward” and
“outward” flange-lip motions, respectively – this configuration is depicted in Fig. 13(a1). Once more,
again due to the initial geometrical imperfection shape, the deformed configuration is not symmetric,
as attested by the fact that the mid-height cross-section exhibits “outward” flange-lip motions. As was
mentioned in item (vi), the minor-axis flexural displacements point towards the web (see Fig. 14(c4)).
Lastly, Figs. 13(a1)-(a3) show that almost exclusively distortional deformations develop before the
peak load is reached, since the local deformations only emerged at fairly high loads − for instance,
visible flange local deformations can only be observed in Fig. 13(a4), which concerns an applied load
level well above those associated with Figs. 13(a1)-(a3)) – moreover, compare the flange deformed
configuration shown in Fig. 13(a4) with those appearing in Figs. 13(a2)+(a3). This behavioural feature
is in agreement with the theory, since distortional buckling precedes local buckling load for this
particular column (RDL0.80). Finally, note that L-D interaction occurs because the column is locally
and distortionally rather slender − for less slender columns, the failure mode may be pure distortional
(i.e., not exhibit local deformations). Rigorously speaking, this column geometry is associated with
the transition between “secondary local-bifurcation L-D interaction” and “true L-D interaction”.

3.5.1.2 Secondary distortional-bifurcation L-D interaction

The results presented in this sub-section, which are qualitatively similar to those presented previously,
concern columns WS-12 and WS-13 and consist of (i) photos showing column deformed configurations
prior to collapse (Figs. 15(a1)-(a4) and Figs. 16(a1)-(a4)) or at the onset of collapse (Figs. 15(b1)-(b4) and
Figs. 16(b1)-(b4)), and (ii) the recorded column axial shortening and displacements measured by
transducers T1, T4 and T5 (Fig. 17). The observation of these results prompts the following comments:
(i) The equilibrium paths P vs. T5 (Fig. 17(d)), qualitatively similar to those obtained numerically
(e.g., Martins et al. 2018), are almost linear behaviour at the early loading stages, up until the
vicinity of the critical (local) buckling load (26kN – see Table 6). Thereafter, they become non-
linear and associated with considerable post-critical strengths up until P50kN (i.e., close to the
level of the theoretical critical distortional buckling load. For higher applied loads, there is noticeable
strength erosion that happens while the distortional deformations grow fast and continues until the
collapse – note that the failure load is much higher than the theoretical PcrD value (slender
column). The emergence of pronounced distortional deformations (see, for instance, Fig. 17(b)) in
columns where local buckling precedes distortional buckling characterises the “secondary
distortional-bifurcation L-D interaction”.
(ii) The photos displayed in Figs. 15 and 16, concerning deformed configurations either before or at the
onset of collapse, show once again clear distortional (mostly) and local deformations even when local
buckling precedes distortional buckling by a significant amount (RDL2.00) − this characterises L-D
interaction (e.g., Martins et al. 2018). On the other hand, it was observed that, for applied
loads between the theoretical critical local and distortional buckling loads, mostly local
deformations occur in these two columns − they are much more pronounced than those
appearing in the columns addressed in the previous sub-section (unfortunately, there exist no
photos backing/illustrating this assertion). Once distortional buckling is triggered, the distortional
deformations grow rapidly and outweigh their local (and minor-axis flexural) counterparts.

19
 (a1) (a2) (a3) (a4)

(b1) (b2) (b3) (b4)

Figure 15: WSLC-12 column deformed configurations (a) prior to collapse (equilibrium path ascending branch) and (b) at the onset of
collapse (failure mode).

(iii) The WS-12 and WS-13 column failure modes exhibit, once again, a single distortional half-wave
that is akin to the critical distortional buckling mode and involves “inward” flange-lip motions (like
all WSLC columns affected by “true L-D interaction”).
(iv) The equilibrium paths P vs. T4 (Fig. 17(c)) show that significant minor-axis flexural
displacements emerge, mainly after the development of considerable distortional deformations −
they pointing towards the web, like in all the WSLC columns addressed in the previous sub-section.
(v) Unlike in most of the WSLC columns addressed in the previous sub-section, the plastic hinges
now developed in the mid-height region (see, for instance, Figs. 16(b1)-(b2) and also Fig. 19),
which is in close agreement with available numerical simulations (e.g., Martins et al. 2016).
These plastic hinge locations are very similar to those exhibited by the failure modes of columns
buckling and failing in pure distortional modes (involving one half-wave) − this fact is responsible
for the drastic strength erosion associated with this type of L-D interaction. Indeed, if the occurrence
of L-D interaction could be fully prevented, the columns would collapse in pure local modes, which
have a completely different nature − i.e., L-D interaction causes a clear change in failure mode nature.

3.5.2 Failure loads

The values of the experimental failure loads (PExp) obtained in this investigation for the 31 WFSLC and
WSLC column tested are given in Table 5. This table also shows (i) the maximum displacements at or
just before the peak load, namely the axial shortening () and the flange-lip vertical displacement (D),
and (ii) the observed failure mode nature (“L” and “D” stand for local and distortional deformation,

20
 (a1) (a2) (a3) (a4)

(b1) (b2) (b3) (b4)

Figure 16: WSLC-13 column deformed configurations (a) prior to collapse (ascending branch) and (b) at the onset of collapse (failure mode).
WS-12 P (kN) WS-12 P (kN) WS-12 P (kN) WS-12
80 80 80
70 70 70

60 60 60

50 50 50

40 40 40

30 30 30

20 20 20

10 10 10

0 0 0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 0 10 20 30 40 50 0 3 6 9 12 15 0 5 10 15 20 25 30
Axial shortening (mm) Displacement T1 (mm) Displacement T4 (mm) Displacement T5 (mm)
(a1) (b1) (c1) (d1)
WS-13 P (kN) WS-13 P (kN) WS-13 P (kN) WS-13
80 80 80
70 70 70

60 60 60

50 50 50

40 40 40

30 30 30

20 20 20

10 10 10

0 0 0
0.0 1.0 2.0 3.0 4.0 5.0 0 10 20 30 40 0 2 4 6 8 10 0 5 10 15 20 25
Axial shortening (mm) Displacement T1 (mm) Displacement T4 (mm) Displacement T5 (mm)
(a2) (b2) (c2) (d2)
Figure 17: Experimental equilibrium paths: applied load P vs. (a) axial shortening, and displacements (b) T1, (c) T4 and
(d) T5, concerning specimens (1) WS-12 and (2) WS-13.

21
respectively). As mentioned earlier, all specimens failed in L-D interactive modes and, with one single
exception, exhibited one distortional half-wave − the exception was specimen WS-1, which exhibited two
distortional half-waves. Out of the 30 specimens exhibiting one distortional half-wave, those exhibiting
failure modes involving “inward” or “outward” flange-lip motions are identified in Table 5 by the
designations “L-D–” and “L-D+”, respectively − the numbers of distortional half-waves are given between
brackets. The failure loads of the repeated specimens (WFS-5R, WFS-6R, WFS-9R, WS-5R and WS-6R)
differed only by 1.5, 1.0, 1.3, 0.2 and 1.8% from their “original” counterparts, thus indicating an excellent
reliability of the test set-up and procedure. In addition, the “original” and repeated specimens showed
similar deformed configurations at the onset of collapse − the small failure load differences are naturally
attributed to the distinct initial geometrical imperfections. Finally, Figs. 18 and 19 show photos of the
whole sets of WFSLC and WSLC specimens after the completion of the experimental procedure, showing
their deformed configurations once the applied load has been fully removed.

Table 5: Column experimental failure loads, corresponding displacements and observed failure mode natures.
Specimen PExp  D Failure mode nature
(kN) (mm) (mm)
WFS-1 86.1 4.6 44.5 L+D(1)+
WFS-2 86.5 4.1 44.7 L+D(1)+
WFS-3 85.9 4.1 40.5 L+D(1)+
WFS-4 115.8 4.3 39.1 L+D(1)+
WFS-5 110.9 4.7 45.1 L+D(1)+
WFS-5R 112.6 5.1 44.0 L+D(1)+
WFS-6 86.5 4.4 39.5 L+D(1)+
WFS-6R 85.6 3.8 39.1 L+D(1)+
WFS-7 71.2 4.3 -46.8 L+D(1)–
WFS-8 86.5 4.0 39.4 L+D(1)+
WFS-9 88.7 3.5 30.8 L+D(1)+
WFS-9R 89.9 3.9 32.3 L+D(1)+
WFS-10 86.4 3.5 35.0 L+D(1)+
WFS-11 87.7 3.8 36.0 L+D(1)+
WFS-12 80.9 3.9 -44.1 L+D(1)–
WFS-13 84.3 4.1 38.7 L+D(1)+
WS-1 112.7 5.3 -15.0 L+D(2)
WS-2 117.9 6.4 -9.8 L+D(1)–
WS-3 120.7 4.2 -17.3 L+D(1)–
WS-4 122.4 3.5 -24.2 L+D(1)–
WS-5 119.6 4.8 -40.9 L+D(1)–
WS-5R 119.8 4.9 -25.3 L+D(1)–
WS-6 120.5 4.4 -30.4 L+D(1)–
WS-6R 118.4 4.2 -37.7 L+D(1)–
WS-7 120.3 3.6 -30.1 L+D(1)–
WS-8 118.1 5.0 -33.5 L+D(1)–
WS-9 86.7 5.4 -44.7 L+D(1)–
WS-10 86.0 5.6 -44.3 L+D(1)–
WS-11 68.9 5.2 -46.9 L+D(1)–
WS-12 65.6 4.0 -40.0 L+D(1)–
WS-13 66.7 3.9 -29.4 L+D(1)–

22
 Figure 18: Deformed configurations of all WFSLC specimens after the full removal of the applied load.

Figure 19: Deformed configurations of all WSLC specimens after the full removal of the applied load.

4. Direct Strength Method (DSM) design

This section presents the failure load estimates provided by the currently codified DSM-based column
design curves, namely the local and distortional design curves, for the tested specimens, and compares
them with the experimental failure loads obtained during this investigation. Since all specimens failed in
L-D interactive modes, it makes no sense to compare their failure loads with the predictions of the global
or local-global interactive design curves. Table 6 presents the (i) critical local and distortional buckling
loads, obtained with GBTUL (Bebiano et al. 2018) on the basis of the measured geometries (Table 2) and
determined elastic properties (Table 3), and (ii) local and distortional slenderness values, both required to
obtain the DSM failure load predictions associated with distortional (PnD) and local (PnL) collapses. In
addition, Figs. 20(a)-(b) show the PExp/PnD vs. λD and PExp/PnL vs. λL plots concerning all the experimental
failure loads obtained in this work. The observation of these results leads to the following comments:

23
(i) The comparison between the values given in Tables 1 and 6 shows that the critical buckling load
ratios (RDL) of the planned and tested specimens differ mainly in the critical distortional buckling
loads − the latter are a bit lower, particularly for the WFSLC columns. This is due to a combination
of less wide lips (compared to the designed ones) and less efficient web-stiffeners, since their
horizontal (parallel to the flanges) depth is, in most cases, lower than 15mm (the designed value).
(ii) Fig. 20 shows clearly that the currently codified local and distortional DSM design curves are unable
to provide acceptable failure load predictions for columns undergoing L-D interaction, since they do
not capture the associated failure load erosion − all column failure loads are overestimated and the
overestimation is more pronounced for the slender columns. Moreover, the DSM distortional design
curve naturally outperforms its local counterpart, as attested by comparing their statistical indications
given in Fig. 20(a)+(b) − recall that it predicts lower failure loads (for λL=λD). This confirms previous
findings (e.g., Martins et al. 2017a) and demonstrates, once more, that it is absolutely
indispensable to equip the CFS technical/scientific community with efficient (safe, accurate and
reliable) DSM-based design rules that are able to capture adequately the failure load erosion
stemming from L-D interaction.

Table 6: Tested column specimens (i) geometries, (ii) failure loads, (iii) critical (local and distortional) buckling loads and
(iv) local and distortional slenderness values.
Specimen D B bl t* L fy Py PExp PcrL PcrD RDL λL λD
(mm) (mm) (mm) (mm) (mm) (MPa) (kN) (kN) (kN) (kN) (-) (-) (-)
WS-1 279.6 150.5 15.7 1.513 2501 530 514.0 112.7 89.7 72.8(2) 0.81 2.39 2.66
WS-2 278.7 151.1 16.1 1.515 2201 530 515.8 117.9 90.0 81.9(1) 0.91 2.39 2.51
WS-3 220.6 129.8 15.5 1.513 1701 530 427.8 120.7 103.0 94.4(1) 0.92 2.04 2.13
WS-4 222.3 129.6 16.0 1.522 1502 530 435.3 122.4 106.3 107.3(1) 1.01 2.02 2.01
WS-5 261.9 150.3 16.4 1.512 1803 530 499.0 119.6 88.4 86.7(1) 0.98 2.38 2.40
WS-5R 262.7 149.9 16.3 1.520 1801 530 501.4 119.8 90.2 87.6(1) 0.97 2.36 2.39
WS-6 258.8 150.6 16.3 1.513 1552 530 496.9 120.5 88.0 94.0(1) 1.07 2.38 2.30
WS-6R 259.5 149.8 15.6 1.508 1554 530 491.5 118.4 87.1 85.8(1) 0.98 2.38 2.39
WS-7 269.1 160.3 16.0 1.501 1500 530 516.6 120.3 79.3 88.1(1) 1.11 2.55 2.42
WS-8 269.0 160.4 15.7 1.511 1651 530 520.4 118.1 80.7 82.1(1) 1.02 2.54 2.52
WS-9 274.5 170.4 16.7 1.215 1653 560 463.1 86.7 39.3 59.1(1) 1.50 3.43 2.80
WS-10 272.7 171.6 16.4 1.219 1955 560 465.0 86.0 39.1 50.7(1) 1.29 3.45 3.03
WS-11 252.1 151.2 16.5 1.013 1753 600 380.4 68.9 26.7 45.4(1) 1.70 3.78 2.90
WS-12 253.3 151.6 16.3 1.011 1652 600 379.5 65.6 26.3 45.6(1) 1.73 3.80 2.88
WS-13 255.5 150.9 15.8 1.012 1350 600 378.3 66.7 26.5 49.9(1) 1.88 3.78 2.75
WFS-1 252.4 202.2 15.9 0.991 1752 600 453.5 86.1 62.9 47.5(1) 0.76 2.69 3.09
WFS-2 252.5 201.5 15.8 1.003 1652 600 462.3 86.5 65.0 55.8(1) 0.86 2.67 2.88
WFS-3 252.2 201.6 15.9 1.002 1503 600 457.4 85.9 65.6 60.4(1) 0.92 2.64 2.75
WFS-4 250.1 220.5 15.8 1.209 1254 560 533.1 115.8 118.7 106.0(1) 0.89 2.12 2.24
WFS-5 271.4 241.2 15.0 1.218 1299 560 571.6 110.9 110.2 82.8(1) 0.75 2.28 2.63
WFS-5R 273.8 240.2 15.0 1.220 1299 560 574.9 112.6 108.6 85.8(1) 0.79 2.30 2.59
WFS-6 270.6 182.6 16.2 1.006 1702 600 443.1 86.5 55.7 49.2(1) 0.88 2.82 3.00
WFS-6R 273.3 181.4 15.8 1.009 1703 600 450.5 85.6 55.3 54.3(1) 0.98 2.85 2.88
WFS-7 283.8 193.2 17.2 0.994 2005 600 458.5 71.2 50.9 40.3(1) 0.79 3.00 3.37
WFS-8 252.5 222.0 15.5 0.998 1355 600 479.8 86.5 65.6 63.7(1) 0.97 2.70 2.74
WFS-9 273.0 182.7 17.7 0.994 1398 600 443.3 88.7 53.2 70.9(1) 1.33 2.89 2.50
WFS-9R 271.5 182.1 16.4 1.010 1401 600 452.2 89.9 56.0 76.2(1) 1.36 2.84 2.44
WFS-10 270.8 182.6 15.6 0.998 1504 600 446.1 86.4 55.5 64.1(1) 1.16 2.84 2.64
WFS-11 282.6 191.6 17.5 1.000 1501 600 464.9 87.7 53.0 68.6(1) 1.29 2.96 2.60
WFS-12 300.5 202.0 16.9 0.999 1654 600 490.9 80.9 48.0 60.0(1) 1.25 3.20 2.86
WFS-13 302.6 201.6 16.7 1.008 1502 600 489.6 84.3 48.4 58.4(1) 1.22 3.19 2.90

24
 PExp /PnD PExp /PnL
1.00 Mean=0.48 WSLC
1.00 Mean=0.34 WSLC
St. Dev.=0.14 St. Dev.=0.09
WFSLC WFSLC
Min=0.25 Min=0.21
0.75 Max=0.70 0.75 Max=0.50

0.50 0.50

0.25 0.25

0.00
D 0.00
L
2.0 2.5 3.0 3.5 2.0 2.5 3.0 3.5 4.0
(a) 3.5 (b)
Figure 20: Plots of (a) PExp/PnD vs. D and (b) PExp/PnL vs. L of all WSLC and WFSLC specimens tested.

In view of what was mentioned in item (ii) above, the authors are currently working on the
development and validation of a DSM-based design approach for CFS columns affected by L-D
interaction that (i) predicts safely, accurately and reliably the whole set of available experimental and
numerical failure loads, and (ii) hopefully will be a strong candidate for codification in the near future.
The work under way, which will be reported soon, (i) includes the performance of fresh numerical
simulations, prompted by the experimental results reported in this work and using a shell finite element
model validated against them, and (ii) takes into account the currently available (preliminary) DSM-
based design proposals (Silvestre et al. 2012, Martins et al. 2017a, Matsubara et al. 2019).

4. Conclusions
This work reported an experimental investigation, planned at the University of Lisbon and carried out at
The University of Hong Kong, dealing with the structural behaviour and ultimate strength of cold-formed
steel web-stiffened (WSLC) and web-flange-stiffened lipped channel (WFSLC) columns undergoing
several levels of local-distortional (L-D) interaction. After addressing the selection of the column
specimen geometries, which ensured different critical distortional-to-local buckling load ratios, the paper
described (i) the mechanical material properties of the specimens tested, obtained through tensile coupon
tests, and (ii) the test set-up and the procedures related to the test performance, including the location of
the displacement measurements and the initial geometrical imperfections recorded. Then, attention was
turned to the presentation and discussion of the results obtained from the 31 column tests, namely (i) non-
linear load-displacement equilibrium paths, (ii) photos showing deformed configurations along those
paths (including the failure modes) and (iii) failure loads. Lastly, the paper compared the experimental
failure loads obtained in this investigation with their estimates provided by the currently codified DSM
(Direct Strength Method) local and distortional design curves.
This investigation provided clear experimental evidence about the occurrence of L-D interaction in
WSLC and WFSLC slender columns for a fairly large range of critical distortional-to-local buckling
load ratios, comprised between 0.75 and 1.88. In addition, it (i) enabled the acquisition of substantial
experimental failure load data (the main purpose of this investigation) and (ii) provided experimental
validation/confirmation of several findings obtained in the recent past, on the basis of numerical
results (similar load-displacement equilibrium paths and failure modes) and (iii) showed that L-D
interaction is mostly governed by distortional deformations (even when local buckling precedes its
distortional counterpart). The comparison between the experimental failure loads and their predictions
provided by the currently codified local and distortional DSM design curves confirmed the high impact

25
of L-D interaction, in the sense that it leads to substantial failure load erosion, and also the inability of the
above design curves in capturing this impact. It also demonstrated, once more, the need for design rules
able to handle properly column L-D interactive failures. The authors are currently working on fulfilling
this need, by developing an efficient DSM-based design approach intended for cold-formed steel
columns affected by L-D interaction that can be proposed for codification − the fruits bore by this research
effort will be reported soon. In this context, the column failure loads acquired in this work, together
with those already available (either experimental or numerical), will certainly play a pivotal role in paving
the way towards the sought codification − the ultimate goal of this ongoing investigation.

Acknowledgments
The authors thankfully acknowledge the support of M Metal Pte Ltd in Singapore, for supplying the cold-
formed steel column specimens to be tested, and the assistance of Mr. Cheuk-Hei Chow during the
performance of the experimental test campaign at The University of Hong Kong.

References
ASTM-E8M (2016). “Standard test methods for tension testing of metallic materials”, ASTM E8M-16, West Conshohocken,
PA, USA.
Bebiano, R., Camotim, D., Gonçalves, R. (2018). “GBTUL 2.0 − a second-generation code for the GBT-based buckling and
vibration analysis of thin-walled members”, Thin-Walled Structures, 124(March), 235-257.
Camotim, D., Dinis, P.B., Martins, A.D. (2016). “Direct Strength Method (DSM) − a general approach for the design of cold-
formed Steel structures”, Recent Trends in Cold-Formed Steel Construction, C. Yu (ed.), Woodhead Publishing (Series in
Civil and Structural Engineering), Amsterdam, 69-105.
Camotim, D., Dinis, P.B., Martins, A.D., Young, B. (2018). “Review: Interactive behaviour, failure and DSM design of cold-
formed steel members prone to distortional buckling”, Thin-Walled Structures, 128(July), 12-42.
Chen, J., Chen, M.-T., Young, B. (2019). “Compression tests of cold-formed steel C- and Z-sections with different
stiffeners”, Journal of Structural Engineering (ASCE), 145(5), 04019022 (10 pages).
Dinis, P.B., Young, B., Camotim, D. (2014). “Local-distortional interaction in cold-formed steel rack-section columns”,
Thin-Walled Structures, 81(August), 185-194.
Dinis, P.B., Camotim, D. (2015). “Cold-formed steel columns undergoing local-distortional coupling: behaviour and direct
strength prediction against interactive failure”, Computers & Structures, 147(January), 181-208.
He, Z., Zhou, X., Liu, Z., Chen, M. (2014). “Post-buckling behaviour and DSM design of web-stiffened lipped channel
columns with distortional and local mode interaction”, Thin-Walled Structures, 84(November), 189-203.
Huang, Y., Young, B. (2014). “The art of coupon tests”, Journal of Constructional Steel Research, 96(May), 159-175.
Kwon, Y.B., Hancock, G.J. (1992). “Tests of cold-formed channels with local and distortional buckling”, Journal of Structural
Engineering (ASCE), 118(7), 1786-1803.
Kwon, Y.B., Kim, B.S., Hancock, G.J. (2009). “Compression tests of high strength cold-formed steel channel with buckling
interaction”, Journal of Constructional Steel Research, 65(2), 278-289.
Kwon, Y.B., Kim, N.K., Kim, B.S. (2005). “A study on the direct strength method for compression members undergoing
mixed mode buckling”, Proceedings of 3rd International Symposium on Steel Structures (ISSS’05 − Seoul 10-11/03),
108-119.
Loughlan, J., Yidris, N., Jones, K. (2012). “The failure of thin-walled lipped channel compression members due to coupled local-
distortional interactions and material yielding”, Thin-Walled Structures, 61(December), 14-21.
Martins, A.D., Dinis, P.B., Camotim, D., Providência P. (2015). “On the relevance of local-distortional interaction effects in the
behaviour and design of cold-formed steel columns”, Computers & Structures, 160(November), 57-89.
Martins, A.D., Dinis, P.B., Camotim, D. (2016). “On the influence of local-distortional interaction in the behaviour and design of
cold-formed steel web-stiffened lipped channel columns”, Thin-Walled Structures, 101(April), 181-204.
Martins, A.D., Camotim, D., Dinis, P.B. (2017a). “On the direct strength design of cold-formed steel columns failing in local-
distortional interactive modes”, Thin-Walled Structures, 120(November), 432-445.
Martins, A.D., Camotim, D., Dinis, P.B. (2017b). “Behaviour and DSM design of stiffened lipped channel columns undergoing
local-distortional interaction”, Journal of Constructional Steel Research, 128(January), 99-118.
Martins, A.D., Camotim, D., Gonçalves, R., Dinis P.B. (2018). “On the mechanics of local-distortional interaction in thin-walled
lipped channel columns”, Thin-Walled Structures, 125(April), 187-202.

26
Matsubara, G.Y., Batista, E.M., Salles, G.C. (2019). “Lipped channel cold-formed steel columns under local-distortional
buckling mode interaction”, Thin-Walled Structures, 137(April), 251-270.
Schafer, B.W. (2008). “Review: the direct strength method of cold-formed steel member design”, Journal of
Constructional Steel Research, 64(7-8), 766-778.
Schafer, B.W. (2019). “Advances in the Direct Strength Method of cold-formed steel design”, Thin-Walled Structures,
140(July), 533-541.
Silvestre, N., Camotim, D., Dinis, P.B. (2012). “Post-buckling behaviour and direct strength design of lipped channel
columns experiencing local/distortional interaction”, Journal of Constructional Steel Research, 73(June), 12-30.
Yang, D., Hancock, G.J. (2004). “Compression tests of high strength steel channel columns with interaction between local
and distortional buckling”, Journal of Structural Engineering (ASCE), 130(12), 1954-1963.
Yap, D.C.Y., Hancock, G.J. (2008). “Experimental study of complex high-strength cold-formed cross-shaped steel section”,
Journal of Structural Engineering (ASCE), 134(8), 1322-1333.
Yap, D.C.Y., Hancock, G.J. (2011). “Experimental study of high strength cold-formed stiffened-web C-sections in
compression”, Journal of Structural Engineering (ASCE), 137(2), 162-172.
Young, B., Rasmussen, K.J.R. (1998). “Design of lipped channel columns”, Journal of Structural Engineering (ASCE),
124(2), 140-148.
Young, B., Rasmussen, K.J.R. (1999). “Shift of effective centroid in channel columns”, Journal of Structural Engineering
(ASCE), 125(5), 524-531.
Young, B., Silvestre, N., Camotim, D. (2013). “Cold-formed steel lipped channel columns influenced by local-
distortional interaction: strength and DSM design”, Journal of Structural Engineering (ASCE), 139(6), 1059-
1074.

27